Exelixis, Inc. (Nasdaq: EXEL) reported that the final analysis of its second dual primary endpoint in the Phase III STELLAR-303 trial did not reach statistical significance, revealing a split result that creates uncertainty around how FDA will interpret the totality of evidence ahead of a December 2026 decision date.
The trial, which evaluated zanzalintinib in combination with Genentech's atezolizumab versus Bayer's Stivarga (regorafenib) in previously treated non-MSI-high metastatic colorectal cancer, carried two co-primary overall survival endpoints: one in the full intention-to-treat population and one in patients without active liver metastases at baseline. Exelixis previously reported that the ITT endpoint was met, with data presented at the 2025 European Society for Medical Oncology Congress and published in The Lancet. The non-liver metastases subgroup, however, produced a hazard ratio of 0.83 (95% CI: 0.66–1.05; P=0.1185) — a numerical trend in the right direction but short of the pre-specified threshold for statistical significance. Median OS was 15.9 months with zanzalintinib plus atezolizumab versus 12.7 months with regorafenib in that subgroup.
The regulatory question
The FDA accepted Exelixis's New Drug Application in February 2026, with a PDUFA date of December 3, 2026. The submission covers the broader previously treated mCRC population, not the NLM subgroup alone. That the ITT endpoint was met provides the foundation for the application. Whether the failure of the NLM subgroup to achieve statistical significance creates a labeling complication, or prompts agency questions about the consistency of benefit across patient subsets, is the central uncertainty now facing the program.
The NLM subgroup result is not trivial in clinical terms. Liver metastases in colorectal cancer are associated with an immunosuppressive microenvironment that substantially limits checkpoint inhibitor activity — a well-documented phenomenon that has shaped trial design in this space. The hypothesis underpinning the NLM subgroup endpoint was that patients without liver involvement would derive the greatest benefit from immune checkpoint inhibition. The observed hazard ratio of 0.83 is directionally consistent with that hypothesis, but the confidence interval crossing 1.0 means the data do not statistically confirm it.
Zanzalintinib is an oral inhibitor of the TAM kinases (TYRO3, AXL, MER), MET, and VEGF receptors. The mechanistic rationale for combining it with atezolizumab rests on its potential to reduce immune evasion through TAM kinase blockade and to suppress angiogenesis, thereby creating a more permissive tumor microenvironment for checkpoint inhibitor activity. The safety profile in the NLM subgroup was consistent with the ITT population, with no new signals identified.